Problem in ASR environment.

[noob3004]

Thank you for your outstanding work. I am trying it, I can split audio normally in the new environment created by Readme, but I cannot use ASR. There are something wrong in mindscore?

KeyError: 'funasr-pipeline is not in the pipelines registry group auto-speech-recognition. Please make sure the correct version of ModelScope library is used.'

I would be extremely grateful if you could answer.

[Randolph87cb]

试试用git源安装？
funasr 1.0.0
modelscope 1.9.4

https://github.com/modelscope/modelscope.git
https://github.com/alibaba/FunASR.git

git clone xxx
cd xxx
pip install -e ./

[noob3004]

试试用git源安装？ funasr 1.0.0 d:\workspace\gpt-sovits\venv\scripts\funasr modelscope 1.9.4 d:\workspace\gpt-sovits\venv\scripts\modelscope\package

https://github.com/modelscope/modelscope.git https://github.com/alibaba/FunASR.git

git clone xxx cd xxx pip install -e ./

感谢回复，估计就是作者所言版本不一接口有变化了。

[mrhan1993]

如果没有用，可以试下我的办法：
将 tools\damo_asr\cmd-asr.py 文件中的

inference_pipeline = pipeline(
    task=Tasks.auto_speech_recognition,
    model='tools/damo_asr/models/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch',
    vad_model='tools/damo_asr/models/speech_fsmn_vad_zh-cn-16k-common-pytorch',
    punc_model='tools/damo_asr/models/punc_ct-transformer_zh-cn-common-vocab272727-pytorch',
)

改成

inference_pipeline = pipeline(
    task=Tasks.auto_speech_recognition,
    model='damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch',
    vad_model='damo/speech_fsmn_vad_zh-cn-16k-common-pytorch',
    punc_model='damo/punc_ct-transformer_zh-cn-common-vocab272727-pytorch',
)

根据我的研究，原因是 git clone 的那几个模型的配置文件和它自动下载的内容不一样

[RVC-Boss]

@mrhan1993 国外用户下载modelscope速度咋样

这个写法是整合包写法，照顾网不好的用户的，老版readme也写了下载方式https://github.com/RVC-Boss/GPT-SoVITS/tree/3b1ba91e30c7aff204191fee10bff2719b0adb1e

[RVC-Boss]

我把readme完善了。当然大家也可以按照@mrhan1993提供的方法自己手工修改这个路径自动下载modelscope模型。

[mrhan1993]

@RVC-Boss 国外 modelscope 这还真不知道。不过我手动下载模型就会报这个 KeyError

[lyris]

如果没有用，可以试下我的办法： 将 tools\damo_asr\cmd-asr.py 文件中的

inference_pipeline = pipeline(
    task=Tasks.auto_speech_recognition,
    model='tools/damo_asr/models/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch',
    vad_model='tools/damo_asr/models/speech_fsmn_vad_zh-cn-16k-common-pytorch',
    punc_model='tools/damo_asr/models/punc_ct-transformer_zh-cn-common-vocab272727-pytorch',
)

改成

inference_pipeline = pipeline(
    task=Tasks.auto_speech_recognition,
    model='damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch',
    vad_model='damo/speech_fsmn_vad_zh-cn-16k-common-pytorch',
    punc_model='damo/punc_ct-transformer_zh-cn-common-vocab272727-pytorch',
)

根据我的研究，原因是 git clone 的那几个模型的配置文件和它自动下载的内容不一样

用
funasr==1.0.0 (commit 37d7764ecf0e8cc1a14f59b8b9cd1c914da8b00 origin/funasr1.0 Sun Jan 21 21:06:52 2024 +0800)
modelscope==1.11.0
改 tools\damo_asr\cmd-asr.py 成自动下载仍然报
KeyError: 'funasr-pipeline is not in the pipelines registry group auto-speech-recognition. Please make sure the correct version of ModelScope library is used.'

如果modelscope==1.10.0 错误就变成
ImportError: AutomaticSpeechRecognitionPipeline: cannot import name 'asr_inference_launch' from 'funasr.bin'了
modelscope funasr 好混乱，到目前为止还没有成功

[mrhan1993]

如果没有用，可以试下我的办法： 将 tools\damo_asr\cmd-asr.py 文件中的

inference_pipeline = pipeline(
    task=Tasks.auto_speech_recognition,
    model='tools/damo_asr/models/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch',
    vad_model='tools/damo_asr/models/speech_fsmn_vad_zh-cn-16k-common-pytorch',
    punc_model='tools/damo_asr/models/punc_ct-transformer_zh-cn-common-vocab272727-pytorch',
)

改成

inference_pipeline = pipeline(
    task=Tasks.auto_speech_recognition,
    model='damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch',
    vad_model='damo/speech_fsmn_vad_zh-cn-16k-common-pytorch',
    punc_model='damo/punc_ct-transformer_zh-cn-common-vocab272727-pytorch',
)

根据我的研究，原因是 git clone 的那几个模型的配置文件和它自动下载的内容不一样

用 funasr==1.0.0 (commit 37d7764ecf0e8cc1a14f59b8b9cd1c914da8b00 origin/funasr1.0 Sun Jan 21 21:06:52 2024 +0800) modelscope==1.11.0 改 tools\damo_asr\cmd-asr.py 成自动下载仍然报 KeyError: 'funasr-pipeline is not in the pipelines registry group auto-speech-recognition. Please make sure the correct version of ModelScope library is used.'

如果modelscope==1.10.0 错误就变成 ImportError: AutomaticSpeechRecognitionPipeline: cannot import name 'asr_inference_launch' from 'funasr.bin'了 modelscope funasr 好混乱，到目前为止还没有成功

加群下整合包吧。这个项目环境报错有点莫名其妙的感觉。

[lakerwe]

如果没有用，可以试下我的办法： 将 tools\damo_asr\cmd-asr.py 文件中的

inference_pipeline = pipeline(
    task=Tasks.auto_speech_recognition,
    model='tools/damo_asr/models/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch',
    vad_model='tools/damo_asr/models/speech_fsmn_vad_zh-cn-16k-common-pytorch',
    punc_model='tools/damo_asr/models/punc_ct-transformer_zh-cn-common-vocab272727-pytorch',
)

改成

inference_pipeline = pipeline(
    task=Tasks.auto_speech_recognition,
    model='damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch',
    vad_model='damo/speech_fsmn_vad_zh-cn-16k-common-pytorch',
    punc_model='damo/punc_ct-transformer_zh-cn-common-vocab272727-pytorch',
)

根据我的研究，原因是 git clone 的那几个模型的配置文件和它自动下载的内容不一样

用 funasr==1.0.0 (commit 37d7764ecf0e8cc1a14f59b8b9cd1c914da8b00 origin/funasr1.0 Sun Jan 21 21:06:52 2024 +0800) modelscope==1.11.0 改 tools\damo_asr\cmd-asr.py 成自动下载仍然报 KeyError: 'funasr-pipeline is not in the pipelines registry group auto-speech-recognition. Please make sure the correct version of ModelScope library is used.'
如果modelscope==1.10.0 错误就变成 ImportError: AutomaticSpeechRecognitionPipeline: cannot import name 'asr_inference_launch' from 'funasr.bin'了 modelscope funasr 好混乱，到目前为止还没有成功

加群下整合包吧。这个项目环境报错有点莫名其妙的感觉。

请问群在哪里加呢？

[mrhan1993]

@lakerwe B站视频置顶评论里有 BV12g4y1m7Uw

[potatoker]

如果没有用，可以试下我的办法： 将 tools\damo_asr\cmd-asr.py 文件中的

inference_pipeline = pipeline(
    task=Tasks.auto_speech_recognition,
    model='tools/damo_asr/models/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch',
    vad_model='tools/damo_asr/models/speech_fsmn_vad_zh-cn-16k-common-pytorch',
    punc_model='tools/damo_asr/models/punc_ct-transformer_zh-cn-common-vocab272727-pytorch',
)

改成

inference_pipeline = pipeline(
    task=Tasks.auto_speech_recognition,
    model='damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch',
    vad_model='damo/speech_fsmn_vad_zh-cn-16k-common-pytorch',
    punc_model='damo/punc_ct-transformer_zh-cn-common-vocab272727-pytorch',
)

根据我的研究，原因是 git clone 的那几个模型的配置文件和它自动下载的内容不一样

请教下，没太看懂这么改能解决这个问题的原因是？
而且现在版本的这段代码是：

path_asr='tools/damo_asr/models/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch'
path_vad='tools/damo_asr/models/speech_fsmn_vad_zh-cn-16k-common-pytorch'
path_punc='tools/damo_asr/models/punc_ct-transformer_zh-cn-common-vocab272727-pytorch'
path_asr=path_asr if os.path.exists(path_asr)else "damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch"
path_vad=path_vad if os.path.exists(path_vad)else "damo/speech_fsmn_vad_zh-cn-16k-common-pytorch"
path_punc=path_punc if os.path.exists(path_punc)else "damo/punc_ct-transformer_zh-cn-common-vocab272727-pytorch"
inference_pipeline = pipeline(
    task=Tasks.auto_speech_recognition,
    model=path_asr,
    vad_model=path_vad,
    punc_model=path_punc,
)

看起来是为了兼容之前放错的目录没什么问题，路径问题上手动下载到这个目录也是按照这个readme操作的。单仍然有KeyError: 'funasr-pipeline is not in the pipelines registry group auto-speech-recognition. Please make sure the correct version of ModelScope library is used.'这个问题，看有的同学说要降级：

pip install modelscope==1.10.0
pip install funasr==0.8.7

我这里降级了但仍然报这个错误。

[veruslana]

如果没有用，可以试下我的办法： 将 tools\damo_asr\cmd-asr.py 文件中的

inference_pipeline = pipeline(
    task=Tasks.auto_speech_recognition,
    model='tools/damo_asr/models/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch',
    vad_model='tools/damo_asr/models/speech_fsmn_vad_zh-cn-16k-common-pytorch',
    punc_model='tools/damo_asr/models/punc_ct-transformer_zh-cn-common-vocab272727-pytorch',
)

改成

inference_pipeline = pipeline(
    task=Tasks.auto_speech_recognition,
    model='damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch',
    vad_model='damo/speech_fsmn_vad_zh-cn-16k-common-pytorch',
    punc_model='damo/punc_ct-transformer_zh-cn-common-vocab272727-pytorch',
)

根据我的研究，原因是 git clone 的那几个模型的配置文件和它自动下载的内容不一样

请教下，没太看懂这么改能解决这个问题的原因是？ 而且现在版本的这段代码是：

path_asr='tools/damo_asr/models/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch'
path_vad='tools/damo_asr/models/speech_fsmn_vad_zh-cn-16k-common-pytorch'
path_punc='tools/damo_asr/models/punc_ct-transformer_zh-cn-common-vocab272727-pytorch'
path_asr=path_asr if os.path.exists(path_asr)else "damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch"
path_vad=path_vad if os.path.exists(path_vad)else "damo/speech_fsmn_vad_zh-cn-16k-common-pytorch"
path_punc=path_punc if os.path.exists(path_punc)else "damo/punc_ct-transformer_zh-cn-common-vocab272727-pytorch"
inference_pipeline = pipeline(
    task=Tasks.auto_speech_recognition,
    model=path_asr,
    vad_model=path_vad,
    punc_model=path_punc,
)

看起来是为了兼容之前放错的目录没什么问题，路径问题上手动下载到这个目录也是按照这个readme操作的。单仍然有KeyError: 'funasr-pipeline is not in the pipelines registry group auto-speech-recognition. Please make sure the correct version of ModelScope library is used.'这个问题，看有的同学说要降级：

pip install modelscope==1.10.0
pip install funasr==0.8.7

我这里降级了但仍然报这个错误。

这边似乎解决了。
mrhan1993的说法是正解。modelscope自动下载的可以用。手动git clone的会遇到问题。

解决方案1：
删除手动下载的文件。代码会自动下载需要的模型。

解决方案2：
用mrhan1993提供的这段代码自动下载模型到cache目录。然后再copy到tools/damo_asr/models/目录下。后面就可以直接本地加载了。

path_asr="damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch"
path_vad="damo/speech_fsmn_vad_zh-cn-16k-common-pytorch"
path_punc="damo/punc_ct-transformer_zh-cn-common-vocab272727-pytorch"

inference_pipeline = pipeline(
task=Tasks.auto_speech_recognition,
model=path_asr,
vad_model=path_vad,
punc_model=path_punc,
)